INTRODUCTION

Type II iodothyronine 5-deiodinase (5D-II)¹ catalyzes the metabolism of T₄ to its active metabolite, T₃, in the brain (1, 2). In astrocytes, 5D-II is a multimeric protein with marked molecular asymmetry and has an apparent molecular mass of ~200 kDa (3). 5D-II-like catalytic activity has also been reported in the pituitary and brown adipose tissue (1); however, the hydrodynamic properties of the responsible protein(s) are unknown. One unique feature of brain 5D-II is the rapid, thyroxine-dependent regulation of catalytic activity observed both in the intact animal and in cell culture (4, 5, 6, 7, 8, 9). Cyclic AMP-stimulated astrocytes express abundant 5D-II catalytic activity (10) and faithfully mimic the thyroid hormone-induced changes seen in vivo. For example, removal of thyroid hormone from the culture medium results in a 10-fold increase in 5D-II activity in cAMP-stimulated astrocytes (7). This increase in enzyme activity results from slowed enzyme inactivation and not from increased enzyme synthesis (7, 11). Acute T₄ replacement leads to a rapid loss of 5D-II activity that is independent of gene transcription or protein synthesis (7), but can be blocked by cytochalasin-induced disruption of the actin cytoskeleton (7, 12).

The alkylating thyroid hormone analog N-bromoacetyl-T₄ (BrAcT₄) has proven invaluable for identifying deiodinase polypeptides (8, 13, 14). Under specific conditions, three to five astrocyte proteins contain >95% of the incorporated BrAcT₄ label, and two of these have been identified. A 55-kDa affinity-labeled protein (p55) is the subunit monomer of protein-disulfide isomerase (15). The 29-kDa affinity-labeled protein (p29) was identified as the substrate-binding protein of 5D-II. The following criteria were used to establish the identity of the p29 protein. (i) An increase in affinity labeling paralleled the cAMP-stimulated increase in 5D-II catalytic activity; (ii) the quantity of BrAcT₄ incorporated was directly proportional to 5D-II activity; (iii) affinity labeling was competitively inhibited by substrates T₄ and rT₃, but not the product T₃; and (iv) the rate of inactivation of 5D-II by the affinity label equaled the rate of BrAcT₄ incorporation into p29 (8, 11, 12). Subsequently, the T₄-dependent increase in 5D-II inactivation was shown to coincide with the migration of p29 from the plasma membrane to the endosomal storage pool, and the rate of loss of enzyme activity equaled that of the internalization of p29 (11). Cytochalasin-mediated depolymerization of the actin cytoskeleton blocked both T₄-dependent enzyme inactivation and p29 internalization (11, 12).

In addition to the well characterized p29 protein of 5D-II, an additional 29-kDa protein was weakly labeled by BrAcT₄ in astrocytes lacking 5D-II catalytic activity (8). Whether this protein was related to p29 or was another T₄-binding protein or an artifact of affinity labeling (16) was not known. We therefore re-examined the conditions of p29 affinity labeling, compared p29 with the 29-kDa polypeptide in unstimulated astrocytes, and evaluated the mechanism of cAMP activation of 5D-II.

EXPERIMENTAL PROCEDURES

Materials

All chemicals were of the highest grade available. Dulbecco's modified Eagle's medium was obtained from Life Technologies, Inc., and supplemented bovine calf serum was from Hyclone Laboratories (Logan, UT). N-Bromoacetyl-L-T₃ and BrAcT₄ were obtained courtesy of Dr. Hans Cahnmann (National Institutes of Health, Bethesda, MD).

Methods

Astrocytes were obtained from 1-day-old neonatal rat cerebral hemispheres as described previously (10). Astrocytes were grown in a humidified atmosphere of 5% CO₂ and 95% air in Dulbecco's modified Eagle's medium containing 15 mM sodium bicarbonate, 33 mM glucose, 1 mM sodium pyruvate, 15 mM HEPES, pH 7.4, 10% (v/v) supplemented bovine calf serum, 50 units/ml penicillin, and 90 µg/ml streptomycin. Cells were passaged weekly and used between passages 1 and 3. For all experiments, unless otherwise noted, maximal 5D-II activity was induced in confluent monolayers by growth in serum-free medium for 24 h, followed by an additional 16 h with 1 mM dibutyryl cAMP and 100 nM hydrocortisone (10).

Affinity Labeling of 5D-II

BrAc[¹²⁵I]T₄ was prepared by radioiodination of N-bromoacetyl-L-T₃ as described previously (13). Cells were affinity-labeled at 37 °C with 1.3 nM BrAc[¹²⁵I]T₄ (specific activity of 2200 Ci/mmol) in buffered Hanks' solution containing 50 mM HEPES, pH 7.4, unless otherwise indicated. After a 20-min incubation, cells were washed free of unincorporated affinity ligand, scraped from the dish, and collected by centrifugation. Cells were resuspended in 10 mM HEPES, pH 7.0, containing 10 mM dithiothreitol, 0.1 mM EDTA, and 0.1 mM phenylmethylsulfonyl fluoride (lysis buffer) and sonicated. Either cell lysates were used directly for SDS-PAGE analysis, or crude microsomal preparations were obtained by centrifuging the lysate for 30 min at 250,000 × g through a 0.8 M sucrose cushion. Microsomes were resuspended by trituration in lysis buffer and used as described below.

Microsomal preparations from affinity-labeled astrocytes were solubilized in 5 mM taurodeoxycholate, and the detergent extracts were clarified by centrifugation for 30 min at 250,000 × g. The resultant supernatant was separated on a 90 × 1.5-cm Sephacryl S-300 column equilibrated in 50 mM NH₄Ac, pH 7.0, containing 1 mM dithiothreitol, 0.1 mM EDTA, and 5 mM taurodeoxycholate at a flow rate of 10 ml/h, and 1-ml fractions were collected. The detergent-soluble extracts from microsomes isolated from 3 × 10⁸ cells were used for each separation, and 100-µl aliquots of selected fractions were analyzed directly by SDS-PAGE. The distribution of p29 was quantitated either by densitometry or by counting the electrophoretically resolved p29 bands in a -counter. The column was standardized using thyroglobulin, -amylase, rabbit IgG, -galactosidase, ovalbumin, and cytochrome c with dextran blue and ³H₂O used for the void volume and total volume, respectively.

Affinity-labeled 29-kDa proteins from cAMP-stimulated and unstimulated astrocytes were isolated from microsomal preparations by SDS-PAGE on 8-14% gradient gels as described previously (3). The 29-kDa proteins were cleaved directly in the gel slices with Staphylococcus aureus V8 protease or cyanogen bromide using a modification of the Cleveland method (see Refs. 3, 17, and 18). Digestion products were separated on a 15% SDS-polyacrylamide gels. Peptide fingerprints were compared by autoradiography.

Cyclic AMP-stimulated astrocytes (~10⁹ cells) were affinity-radiolabeled with BrAc[¹²⁵I]T₄ (1 × 10⁶ cpm/pmol) for 15 min, followed by a 15-min incubation with excess 1 µM BrAcT₄ to quantitatively label the p29 polypeptide (19, 20). Microsomes were prepared by discontinuous sucrose gradient centrifugation, and the membrane preparation was solubilized with 5 mM taurodeoxycholate. BrAc[¹²⁵I]T₄-labeled proteins were isolated by immunoprecipitation with anti-T₄ IgG (12), and the radiolabeled proteins in the immune pellet were separated by preparative SDS-PAGE. The gel fragment containing p29 was homogenized in 125 mM Tris-HCl buffer, pH 6.8, mixed with an equal volume of complete Freund's adjuvant and used to immunize 2.2-kg female New Zealand White rabbits. Rabbits were boosted 3 weeks later using the same preparation of p29 homogenized in 125 mM Tris buffer, pH 6.8, alone. Sera containing anti-p29 antibody were identified by Western blotting against purified p29 and used for immunoprecipitation of solubilized affinity-labeled p29- and 5D-II-containing vesicles. Subcellular localization of 5D-II was determined by immunocytochemistry as detailed in the figure legends.

Affinity-labeled astrocytes grown in serum-free medium were treated with 10 nM T₄ for 30 min to initiate endocytosis of the 5D-II-containing vesicles (11, 12). Cells were scraped from the dish, collected by centrifugation, and then lysed by three freeze-thaw cycles. Cell lysates were centrifuged (805,000 × g_min) through 16% Percoll gradients in 250 mM sucrose in 10 mM HEPES buffer, pH 7, containing 10 mM dithiothreitol and 1 mM EDTA, and 0.5-ml fractions were collected. Fractions containing the endosomes (>80% of internalized 5D-II (12)) were pooled and then incubated for 30 min at 4 °C with magnetic beads (Dynabeads, Dynal, Inc.) coated with purified IgG according to the manufacturer's instructions. Control incubations included magnetic beads coated with bovine serum albumin. The beads were then collected; washed three times with 175 mM sodium phosphate buffer, pH 7.4, containing 5 mg/ml bovine serum albumin; and resuspended in SDS-PAGE sample buffer. Proteins were eluted from the beads by boiling for 5 min and resolved by SDS-PAGE.

Astrocytes were grown on poly-D-lysine-coated coverslips as described previously (12) and treated as indicated in the figure legends. Cells were fixed with 4% paraformaldehyde and permeabilized with Triton X-100 as detailed previously (12). BrAcT₄-labeled proteins and p29 were visualized by indirect immunofluorescence using rabbit anti-T₄ IgG and rabbit anti-p29 IgG, respectively; immune complexes were identified with Texas Red-conjugated anti-rabbit IgG. Cells were imaged by confocal microscopy as described previously (12). Micrographs shown are representative of 30-40 independent fields.

RESULTS

One criterion that established p29 as the substrate-binding subunit of 5D-II was the observation that the quantity of affinity-labeled p29 was directly proportional to 5D-II catalytic activity (8). However, in unstimulated cells lacking 5D-II activity, another 29-kDa protein was also weakly labeled with BrAcT₄. To characterize this latter protein, we established conditions that optimized the BrAcT₄ labeling of this T₄-binding protein. The effect of increasing concentrations of BrAc[¹²⁵I]T₄ on affinity labeling of astrocyte polypeptides is shown in Fig. 1. At low concentrations, BrAcT₄ incorporation into the 29-kDa protein in unstimulated cells was only 15% of that observed for p29 in cAMP-stimulated astrocytes as reported previously (8). At increasing concentrations of affinity ligand, this differential labeling pattern was progressively overcome, and little or no difference in affinity labeling of the 29-kDa protein(s) was observed at concentrations greater than ~2 nM BrAc[¹²⁵I]T₄. These data identify an affinity-labeled 29-kDa protein in astrocytes lacking 5D-II activity and establish the labeling conditions necessary to allow comparisons between this protein and the p29 protein in cAMP-stimulated cells.

To establish the relationship(s) between these two 29-kDa polypeptides, we used limited proteolysis and peptide fingerprinting. Cyclic AMP-stimulated and untreated astrocytes were affinity-labeled with 10 nM BrAcT₄ for 20 min, conditions that effectively label both p29 and the 29-kDa protein in unstimulated astrocytes (see Fig. 1). The 29-kDa proteins were then isolated by SDS-PAGE, fragments were prepared by CNBr cleavage or V8 protease proteolysis, and the peptides were separated by SDS-PAGE. As shown in Fig. 2, the peptide fingerprints of the p29 protein in cAMP-stimulated astrocytes were identical to those obtained for the 29-kDa protein in unstimulated cells, indicating that these two proteins are closely related, if not identical.

One possible explanation for the inability of unstimulated astrocytes to catalyze 5-deiodination is that the substrate-binding subunit (p29) is dissociated from the holoenzyme. Since p29 is part of an ~200-kDa multimeric complex, cAMP-induced protein-protein interactions could result in the formation of a functional enzyme by causing p29 to associate with the other, yet to be identified subunits in the holoenzyme. Microsomal preparations of affinity-labeled proteins from cAMP-stimulated and unstimulated astrocytes were solubilized in taurodeoxycholate and separated by gel filtration on a Sephacryl S-300 column, and the distribution of BrAc[¹²⁵I]T₄-labeled p29 was determined as described under ``Experimental Procedures.'' Shown in Fig. 3 is a representative chromatogram of p29 in unstimulated and cAMP-stimulated glial cells. The p29 subunit in cAMP-stimulated cells was associated with a complex of ~200 kDa, consistent with previous estimates of holoenzyme size (3). In contrast, in unstimulated cells, p29 was associated with a complex that was ~60 kDa smaller than that in cAMP-stimulated cells. This difference in chromatographic behavior of the p29 protein from the two cell populations was consistently observed (Table I). Since cAMP-induced 5D-II catalytic activity requires both transcription and translation (10), these data suggest that cAMP induces the synthesis of an essential activating factor that associates with the inactive 5D-II complex. There was little or no evidence of p29 monomers or dimers in the chromatograms from either the cAMP-stimulated or unstimulated astrocytes, indicating that most, if not all, of p29 is contained in the multimeric holoenzyme.

-labeled 5D-II from stimulated () and unstimulated () astrocytes.

-am

-gal

Table I. Gel filtration analysis of BrAc[125I]T4-labeled p29 from unstimulated and cAMP-stimulated glial cells Run No. Column Vt-V0 p29 (Kav) Unstimulated cAMP-stimulated (ml) 1 I 51.5 0.25 0.20 2 I 0.23 0.19 3 II 68 0.29 0.11 4 II 0.22

To develop a 5D-II-specific immunological probe, anti-p29 antibodies were generated against partially purified p29 from cAMP-stimulated cells as described under ``Experimental Procedures.'' Control studies showed that the anti-p29 antibody was not directed against T₄ per se since anti-p29 IgG failed to immunoprecipitate T₄ (Table II). This eliminated any potential problems of the cross-reactivity of this antibody with other BrAcT₄-labeled proteins.

Table II. Immunoprecipitation of [125I]T4 IgG Immunoprecipitated % Anti-p29 0.9 Preimmune rabbit 3.1 Anti-T4 48.7

The specificity of the anti-p29 antibody is shown in Fig. 4A. As expected, in control immunoprecipitations, anti-T₄ IgG recognized all of the BrAcT₄-labeled proteins (second lane) (11, 12), while anti-p29 IgG preferentially immunoprecipitated the p29 polypeptide (third lane). We then determined if anti-p29 IgG recognized the holoenzyme in its native environment. 5D-II-containing endosomes were prepared from cAMP-stimulated astrocytes by density gradient centrifugation as detailed previously (11, 12) and incubated with immobilized anti-p29 IgG or immobilized normal rabbit IgG as described under ``Experimental Procedures.'' As shown in Fig. 4B (first lane), the total endosomal pool from cAMP-stimulated astrocytes contains three prominent BrAcT₄-labeled proteins (p55, p29, and p18). Previous work has shown that both p55 and p18 are nonspecifically labeled with BrAcT₄, while affinity labeling of p29 is selectively blocked by 5D-II substrates, but not products (8). Immunopurification with anti-p29 IgG-coated beads specifically isolated vesicles containing p29 (third lane), while normal rabbit IgG controls failed to enrich vesicles containing any BrAcT₄-labeled protein (second lane). Consistent with previous reports (11, 12), both the p55 and p18 affinity-labeled proteins were also associated with the immunopurified endosome since the anti-p29 antibody does not recognize either p55 or p18 from detergent-solubilized preparations (see Fig. 4A). These data indicate that the anti-p29 antibody recognizes both detergent-soluble and membrane-bound forms of the p29 subunit of 5D-II.

Immunocytochemical Localization of the p29 Subunit of 5D-II in cAMP-stimulated and Unstimulated Astrocytes

In cAMP-stimulated astrocytes, catalytically active 5D-II is a plasma membrane enzyme, and T₄ regulates 5D-II activity by initiating the translocation of the enzyme from the plasma membrane to the perinuclear space, where it is catalytically inactive (11, 12). Since the p29 subunit is constitutively expressed and part of a multimeric complex in both cAMP-stimulated and unstimulated astrocytes, we determined whether subcellular location played a role in the cAMP induction of catalytically active 5D-II.

A panel of representative photomicrographs of p29 immunoreactivity is shown in Fig. 5. Preimmune serum controls showed no specific immunostaining in cAMP-stimulated, BrAcT₄-labeled astrocytes. Anti-p29 IgG yielded intense staining in cAMP-stimulated, BrAcT₄-labeled astrocytes, presumably over the cell membranes since p29 is an integral membrane protein (8, 12). Interestingly, in the nonaffinity-labeled, cAMP-stimulated astrocytes, the immunoreactivity of p29 was much less than that in BrAcT₄-labeled cells, suggesting that the epitope recognized by anti-p29 IgG requires affinity labeling for maximal exposure.

BrAcT

Since p29 constitutes ~50% of the affinity-labeled protein in astrocytes (8) and the BrAcT₄-labeled proteins are readily recognized by anti-T₄ antibodies, we compared the distributions of anti-T₄ and anti-p29 immunoreactivity in astrocytes. Shown in Fig. 6 are representative confocal micrographs of the effects of T₄ on the subcellular distribution of immunoreactive p29 in BrAcT₄-labeled, cAMP-stimulated astrocytes. As previously reported (11, 12), anti-T₄ IgG shows punctate staining along the cell periphery. Since the nucleus of an astrocyte is elliptical with a long axis of ~8-10 µm and the astrocyte, while polygonal in shape, is ~30 µm in diameter, the nucleus occupies only 25-40% of the cross-sectional area of any given cell, and the remaining intracellular space is filled with other organelles and cell sap. Thus, anti-T₄ IgG immunoreactivity is concentrated over the plasma membrane in the hypothyroid cAMP-stimulated astrocyte, and little, if any, staining is present in the cell interior (Fig. 6A). After 20 min of T₄ treatment (Fig. 6B), the BrAcT₄-labeled protein(s) were lost from the cell membrane and had migrated to the perinuclear space, with occasional punctates observed over the cell nucleus. Since p29 is a membrane-associated protein (8, 11, 12) and, when internalized, is a component of the endosomal vesicle, this pattern of immunoreactivity is consistent with the translocation of affinity-labeled 5D-II from the plasma membrane to the endosomal pool. Parallel astrocyte cultures stained with anti-p29 IgG exhibited identical patterns of rim immunostaining in hypothyroid cells (Fig. 6C) and showed the relocation of the immunoreactive protein to the perinuclear space in the T₄-treated astrocytes (Fig. 6D). These data confirm that anti-p29 IgG recognizes the BrAcT₄-labeled p29 subunit of the 5D-II holoenzyme in intact cells.

treatment on distribution of immunoreactive 5D-II.

Anti-p29 antiserum was then used to determine the subcellular localization of this 5D-II subunit in unstimulated astrocytes grown in thyroid hormone-free medium (Fig. 7). In unstimulated astrocytes, p29 is found in the perinuclear space (P), and little, if any, specific immunoreactivity is associated with the plasma membrane (see arrows). In contrast, in cAMP-stimulated cells, the majority of immunoreactive p29 is found associated with the cell periphery, presumably the plasma membrane (see arrows), and little remains in the perinuclear space. These data suggest that cAMP stimulation results in the translocation of p29 from the perinuclear space to the plasma membrane, coincident with the appearance of catalytically active 5D-II.

Confocal micrographs of 5D-II in stimulated and unstimulated astrocytes.

DBC

DISCUSSION

In this study, we show that p29, the substrate-binding subunit of 5D-II, is constitutively expressed and resides, along with other 5D-II components, in membrane vesicles located in the perinuclear space of unstimulated astrocytes. Cyclic nucleotides induce the appearance of catalytically active 5D-II coincident with translocation of p29 to the plasma membrane. In addition, a cAMP-activating factor(s) is synthesized and becomes associated with the other enzyme components that are stored in vesicles located in the perinuclear space.

The identification of p29 as the substrate-binding subunit of 5D-II was based upon multiple criteria, including a direct proportionality between the quantity of BrAcT₄-labeled p29 and 5D-II activity and the reciprocal relationship between inactivation and p29 labeling. In fact, the rate of enzyme inactivation is identical to the rate of affinity labeling of the p29 subunit (8). At the concentrations of BrAcT₄ used in the early studies (8), minimal background labeling of 29-kDa protein(s) was observed in catalytically inactive cells. In this study, we show that cAMP increases the avidity of p29 for BrAcT₄ by >10-fold. This increase in apparent binding affinity may be due, in part, to the translocation of the enzyme to the cell surface in close proximity to the sites of entry of the affinity ligand and/or to the low K_m for T₄ of the catalytically active enzyme (21).

The presence of the p29 subunit of 5D-II in astrocytes lacking the catalytically active enzyme was confirmed by peptide mapping, gel filtration, and immunocytochemistry. Peptide fingerprints of the p29 proteins in unstimulated and cAMP-stimulated astrocytes were identical. Gel filtration revealed that p29 was a component of a larger membrane-associated complex and was not present as a monomer or dimer in either the unstimulated or cAMP-stimulated astrocyte. These data show that the substrate-binding subunit of 5D-II is constitutively synthesized and integrated into vesicular storage membranes of the cell. Interestingly, cAMP induction of catalytically active 5D-II was accompanied by an increase in the molecular mass of this p29-containing complex, suggesting that an additional cAMP-activating factor(s) is responsible for the generation of active 5D-II.

The molecular events responsible for the cAMP-dependent generation of 5D-II activity in astrocytes are slowly emerging. Cyclic AMP induces expression of an activating factor that associates with the inactive 5D-II complex in storage vesicles and results in an increase in holoenzyme size. This is also accompanied by a marked difference in the subcellular location of the enzyme. Whether the cAMP-induced activation factor is integrated into the 5D-II complex prior to translocation to the plasma membrane or becomes associated with a 5D-II complex that is continually recycled to the plasma membrane is unknown. However, little, if any, immunoreactive p29 is located in the plasma membrane of unstimulated astrocytes, suggesting that recycling of the ``inactive'' 5D-II complex is minor. Thus, the cAMP-activating factor is required for the enzyme to relocate to its site of action at the plasma membrane.

The regulation of 5D-II activity in cultured astrocytes is a complex process that requires cAMP-induced protein synthesis, cAMP-stimulated translocation of an inactive 5D-II complex from intracellular storage pools to the plasma membrane, and microfilament-based endocytosis. In T₄-deficient astrocytes, the biological half-life of the catalytically active enzyme is identical to that of the p29 polypeptide (see accompanying paper (31)). However, in T₄-replete cells, p29 is rapidly internalized by endocytosis, and the biological half-life of catalytically active 5D-II is equal to the rate of p29 endocytosis, while the degradation of the p29 polypeptide remains unchanged under hormone-free or hormone-containing conditions (11). Since the cellular levels of p29 are not regulated by thyroid hormone and p29 is present in astrocytes that do not catalyze 5-deiodination, then regulated expression of this substrate-binding subunit is not the mechanism by which the levels of catalytically active 5D-II are controlled. Moreover, constitutive expression of p29 indicates that the p29 subunit is not the essential cAMP-induced polypeptide. From previous studies, we know that the cAMP-stimulated appearance of 5D-II activity requires the synthesis of an essential protein(s) (10), and it appears that this cAMP-activating factor may be the short-lived protein essential for 5D-II catalysis. The demonstration that an essential cAMP-dependent factor(s) is required to activate the enzyme and target it to its site of action on the plasma membrane identifies one of the other 5D-II subunit polypeptides required for functional 5D-II. In vivo, both the pineal gland (22, 23, 24) and the limbic system (25) in the brain show catecholamine-induced increases in 5D-II activity, suggesting that cAMP is likely to play an important role in regulating the expression of a polypeptide that is essential for 5D-II catalysis, both in vivo and in cell culture.

Brain 5D-II is the largest of the deiodinases, with a calculated molecular mass of 200 kDa (3). Despite the recent cloning of an ~30-kDa frog selenoenzyme with kinetics similar to 5D-II (26), previous biological and biochemical evidence demonstrates that in the mammalian brain, 5D-II is not a selenoenzyme (27, 28, 29, 30). Two of the required subunits of 5D-II are now known, the p29 substrate-binding subunit and an ~60-kDa cyclic AMP-inducible factor. Whether the essential cyclic AMP-inducible factor is an integral part of the enzyme or is merely required to target the enzyme to the plasma membrane remains to be established.